Nvidia’s Big Tech customers might also be its biggest competitive threat
Nvidia’s (NVDA) own Big Tech customers are poised to capture a big chunk of the AI chip market; This dynamic is one that could eventually deepen the chip maker’s profit margins.
Tech giants are making moves to support in-house chip manufacturing businesses. ChatGPT developer OpenAI (OPAI.PVT), a major customer of Nvidia chips by renting cloud space in Microsoft’s (MSFT) and CoreWeave’s (CRWV) data centers, said it will begin designing its own custom chips in partnership with Broadcom.
Meta (META) announced a plan in late September. acquires chip startup Rivos for supporting their own in-house chip work. Amazon (AMZN) said it will launch a massive data center project called Project Rainier this summer, featuring hundreds of thousands of Trainium2 chips to be used by AI developer Anthropic. “on the good road” and analysts said demand for Anthropic’s chips, which are already deployed in data centers, has increased significantly.
While the lion’s share of the AI chip market is dominated by Nvidia’s GPUs (graphics processing units), technology companies led by Alphabet-owned Google (GOOGL, GOOG), Amazon and Microsoft are designing custom chips in partnership with chipmakers Broadcom and Marvell Technology (MRVL).
Analysts explained that the chips are cheaper and better optimized for these companies’ software. While these Big Tech cloud providers don’t sell physical, discrete chips to other companies like Nvidia, the companies do run AI models internally on their own chips, and cloud customers have the option of running AI workloads using these proprietary chips at a lower cost.
Custom chips designed by companies like Google, Amazon, Meta and OpenAI will account for 45% of the AI chip market by 2028, up from 37% in 2024 and 40% in 2025, JPMorgan said in a research note in June. The rest of the chip market is held by GPU manufacturers, namely Nvidia and its rival Advanced Micro Devices (AMD).
The “Magnificent Seven” have good reason to design their own chips.
“The strategic angle here, where all the hyperscalers are looking at custom silicon, is that they don’t want to get stuck behind the NVIDIA monopoly,” Seaport analyst Jay Goldberg said. The incredibly high cost of Nvidia’s AI chips means cloud providers make lower profits Analysts announced that they rented these chips rather than renting their own chips.
“Nvidia now needs to compete with its customers,” Goldberg added.
While tech firms’ custom chips are widely used to run internal AI workloads, Google is reportedly getting started. Sells AI chips physically in the name TPUs (tensor processing units) Its transfer to a cloud provider in September is a move that will allow it to compete directly with companies like Nvidia. DA Davidson analyst Gil Luria estimated that Google’s TPU business, along with its DeepMind AI segment, was worth $900 billion, saying it was “arguably one of Alphabet’s most valuable businesses.”
“Google’s TPUs remain the best alternative to NVIDIA, and the gap between the two has been closing significantly over the last 9-12 months,” Luria wrote in a note to clients in September. he wrote. “[S]”If Google sells its systems to outside customers, the demand will be there, and especially from leading AI labs.”
Alphabet CEO Sundar Pichai unveiled Google’s sixth-generation Trillium TPU at a Google I/O event in 2024. Google released its seventh-generation TPU, Ironwood, in 2025. (AP Photo/Jeff Chiu) ·RELATED PRESS
Overall, Seaport’s Goldberg said he expects “a lot of activity around custom silicon” in 2026, based on conversations he’s had across the AI chip supply chain.
Big Tech companies are at different stages in the evolution of their chip businesses. Google has been developing AI chips called TPUs for more than a decade and is a clear leader among its peers, analysts told Yahoo Finance. Amazon begins its in-house chip journey a year after acquiring Google’s chip startup and launching its first TPU Anapurna Laboratories It was published in 2015 and First Trainium chip of 2020. Meanwhile, Microsoft First dedicated Maia AI chip in 2023 and there is fall behind his peers.
Although custom chips are cheaper to use, AI developers often prefer Nvidia chips because of the software stack that comes with them.
Futurum Group analyst David Nicholson said that although Nvidia is a clear leader today, the tech giants’ custom chip efforts will eventually be reflected in Nvidia’s profits: “Over time, the margins that Nvidia can currently control are diminishing… [it will be] It’s going to cause death by a thousand cuts because you have all these different special silicone accelerators. [chips] They exist because there is such an opportunity.”
When asked about the competitive threat of custom chips last podcast appearanceNvidia CEO Jensen Huang dismissed the concern, saying Nvidia is more than an AI chipmaker because it provides full-scale server systems, not just individual GPUs. While its customers build single chips in most cases, Nvidia designs massive server racks with multiple in-house chips, from Blackwell GPUs to Arm-based central processing units (CPUs) to networking products that allow the chips to talk to each other.
“We are the only company in the world today that produces all the chips within an AI infrastructure,” Huang said on the “BG2” podcast in September.
Nvidia co-founder and CEO Jensen Huang at Computex 2025 at Taipei Music Center in May 2025. (I-Hwa Cheng/AFP via Getty Images) ·I-HWA CHENG via Getty Images
Some analysts also argue that the market for AI chips is so large that there is room for tech firms to expand their custom chip efforts without crowding Nvidia’s market.
Bank of America’s Vivek Arya and Prosecutor Davidson’s Gil Luria said in separate interviews that the growing market share of custom chips “doesn’t matter.”
Arya explained that this is because Nvidia has “managed to continually expand the size of the market.” According to Pitchbook data, Nvidia invested $47 billion worth of venture capital in artificial intelligence companies from 2020 to September this year, investing heavily in the artificial intelligence ecosystem and “neocloud” companies that compete with its own customers.
“Growth and demand are very important,” Luria said. “We’re going to need a lot more calculations and [AI] “Models are becoming more useful, which means the pie will get much bigger in the next few years.”
“Nvidia won’t grow as fast as the market, but they will still be able to grow because the market is growing so fast,” he added.
Concerns about Nvidia are also easing: Not all tech giants embarking on custom chip work will pose an equal threat to Nvidia’s dominance.
“The downside of making your own silicone is that it’s difficult,” Goldberg said. “I think what will happen eventually is that not all of them will succeed.”
StockStory aims to help individual investors beat the market.
Laura Bratton is a reporter for Yahoo Finance. Follow her at Bluesky @laurabratton.bsky.social. Email her at laura.bratton@yahooinc.com.
Click for the latest technology news that will affect the stock market
Read the latest financial and business news from Yahoo Finance
